Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!rice!rice!sun-spots-request
From: del@thrush.mlb.semi.harris.com (Don Lewis)
Newsgroups: comp.sys.sun
Subject: More info on "Why can't I umount ?"
Keywords: Miscellaneous
Message-ID: <1990Aug2.003122.17779@rice.edu>
Date: 31 Jul 90 23:39:55 GMT
Sender: sun-spots-request@rice.edu
Organization: Sun-Spots
Lines: 23
Approved: Sun-Spots@rice.edu
Originator: spots@titan.rice.edu
X-Sun-Spots-Digest: Volume 9, Issue 291, message 4
X-Refs:  Original: 163, Replies: v9n193 v9n206

This problem just bit me again, so I did some digging and it looks like I
found a bug.  The problem is that I am unable to unmount an nfs filesystem
on a client machine.  Umount tells me that the filesystem is busy, but
ofiles does not find any processes referencing the filesystem.

I wrote a program to snoop around in /dev/kmem, and I found that the
mi_refct field in the mntinfo structure for the the filesystem appears to
be corrupted.  The other nfs mounts have small postive reference counts,
but this one has a reference count of -9602!  There is another nfs mount
with a negative reference count (-2921) that is also unmountable.  What is
interesting is that there have been nfs write errors to both servers since
the client was last booted.  In one case the problem was a stale file
handle, in the other case the partition was full.  I am not able to
reproduce the corrupted reference counts by causing these types of write
errors on purpose.

The client is a Sun 4/280 running 4.0.3.  The two servers are a Sun 3/60
running 4.0.3 and a Sun 386i running 4.0.2.  Both the 4/280 and the 3/60
are using the NFS confused file patch.

Don "Truck" Lewis                      Harris Semiconductor
Internet:  del@mlb.semi.harris.com     PO Box 883   MS 62A-028
Phone:     (407) 729-5205              Melbourne, FL  32901